Paper for DEWEK 2015, Category: 9. Lifetime extension monitoring
PRACTICAL EXPERIENCES FROM A LOAD MEASUREMENT CAMPAIGN FOR THE ASSESSMENT OF THE REMAINING SERVICE
LIFE OF WIND TURBINES
Michael Melsheimer**, Dipl.-Ing. René Kamieth*, Prof. Dr.-Ing. Robert Liebich*;
Dr.-Ing. Christoph Heilmann**, Dipl.-Ing. Anke Grunwald**
*Technische Universität Berlin, Fachgebiet Konstruktion und Produktzuverlässigkeit, Institut für Konstruktion,
Mikro- und Medizintechnik, Sekr. H66, Straße des 17. Juni 135, 10623 Berlin, Germany,
Tel.: +49-(0)30-31423603, Fax: +49-(0)30-31426131, rene.kamieth@campus.tu-berlin.de;
**BerlinWind GmbH, Bundesallee 67, 12167 Berlin, Tel.: +49-(0)30-6883337-40, info@berlinwind.com
Summary
After a rapid growth in wind turbine numbers beginning in 1990 in Germany, several thousand turbines now reach
the end of their 20 year design life. Operators often wish to extend the service life of their turbines, but technical
and legal guidelines do not yet specify a clear path. Current methods for the assessment of the remaining service
life include inspection, renewed calculation or a combination of both. However, with these methods, damages
may be detected too late, or fatigue is misjudged, as the endured loads are not realistically considered.
The presented research project proposes a new approach to enhance the above mentioned methods: The loads
on the specific turbine are measured with an efficient load measurement campaign for several weeks to verify its
structural behaviour and loading. The results are combined with additional information of the turbine’s lifetime for
a holistic assessment of the actual lifetime consumption and thus the remaining service life.
A load measurement campaign has been set up in June 2014 in order to test this method, gain practical experience and perform sensitivity studies. Operational data is available for almost the entire lifetime of now 15 years.
Practical challenges include the investigation of classing parameters, as the ones proposed in the appropriate
guidelines often cannot be produced from the available operational data. Evaluations of the data so far have confirmed a highly individual dynamic behaviour, even for turbines of the same or similar type at the same site. The
measurement campaign will be continued to gain additional data and experience.
1. Introduction
Currently, several thousand wind turbines (WT) in
Germany reach the end of their 20-year design life.
For the desired period of continued operation, the
German Building Authority requires an individual
proof of the turbine’s structural integrity and operational safety based on the DIBt guideline (October
2012) [1]. In contrast to DNV-GL [3], the combination of a renewed calculation and a detailed inspection is prescribed as applicable approach.
However, these methods may underestimate the
individual turbine’s actually endured loads since its
structural response depends e.g. on the individual
turbine settings, like intolerable blade angle deviation, resonance issues and/or pre-damage. With a
renewed calculation based on site wind data and
the ideal turbine model, etc., these parameters are
neglected. On the other hand, the inspection detects
damages only when it is quite late for preventive
measures, since the increased loads have been
acting already for a long time at the spot in question.
The presented research project proposes a new
approach to enhance the above mentioned methods: The loads on the specific turbine are measured
with an efficient load measurement campaign for
several weeks to verify its structural behaviour and
loading. The results are combined with additional
information of the turbine’s lifetime for a holistic
assessment of the actual lifetime consumption and
thus the remaining service life.
A load measurement campaign has been set up in
June 2014 in order to test the method, gain practical
experience and perform sensitivity studies. Opera-
tional data is available for almost the entire lifetime
of now 15 years.
2. Approach
2.1 Description of the proposed method
The vibrations of wind turbines are complicated and
have very individual causes. They are dependent on
the site-specific stochastic wind loads, on rotor
imbalances and wear (erosion), and the interaction
of the external forces with the internal structure of
drive train, tower and foundation. In the design
based on standards [3, 5] and guidelines [1, 2], WT
types are extensively investigated by simulation of
operational and extreme load cases for the targeted
wind class The load assumptions are validated by
load measurements according to IEC 61400-13 [2].
For the building approval, it is verified whether the
site's wind conditions are below the wind class limits
for the turbine type, else additional load assessments are necessary. After years of operation, the
turbine has endured some fatigue and will most
likely behave differently compared to the time of
commissioning. A measurement of the current loading situation is then a reliable way of determining a
turbine’s actual dynamic behaviour and hence applied in the approach presented here.
The campaign is set up for efficiency, meaning
reduced duration and number of sensors compared
to an extensive prototype measurement according
to [2]. A goal is to be able to realistically assess the
fatigue condition of the turbine’s supporting structure in an economical manner, even and especially
Paper for DEWEK 2015, Category: 9. Lifetime extension monitoring
for individual and smaller turbines, affected by the
end of their design life right now and where there is
no or limited WT type design data for simulation
available apart from a type approval with some
design load spectra.
2.2 Measurement system
The main requirements for the measurement system are compactness for an economical application
while maintaining a sufficient measurement accuracy for modelling the dynamic behaviour and the
fatigue loading of the wind turbine.
The measurement system for the current measurement campaign consists of 17 sensors (figure 1):
 5 strain gauges, for strain and stress in the tower
top and base,
 5 acceleration sensors, for lateral, axial and torsional accelerations i.e. forces in the nacelle and
the tower top,
 3 turbine data sensors (azimuth position, electrical
power, rotational speed),
 4 meteorological sensors (anemometer and wind
vane on top of the nacelle, ambient air pressure
and temperature below the nacelle).
rotor speed sensor
anemometer,
wind vane
rotor
service life. For this time period, there is at best
operational data as 10 min mean values available,
sometimes even with additional minimum and maximum values. During the short-time measurement
campaign, the operational data is recorded simultaneously with the load data. Classing according to
IEC 61400-13 [2] reduces the data to a more manageable amount and allows a correlation between
operational and load data.
The IEC proposes a two-parameter classing with
mean wind speed vm and turbulence intensity I
(I = σv / vm), which depends on the standard deviation σv and the mean wind speed vm. However,
there is typically no turbulence intensity in the operational data. The search for a reliable substitute
value is one of the challenges in the ongoing research.
3. Current experiences
3.1 Examined turbine
A load measurement runs successfully since June
2014 at a 15 years old 600 kW Pitch turbine with a
65 m steel tower, located in a small wind farm in flat
terrain. The other three turbines are at a distance of
6, 8 and 10 times the rotor diameter, figure 2. The
GL [3] and DIBt [1] guidelines require the consideration of the wake when turbines are closer than 10 D,
which is the case for two turbines here.
nacelle
WT4
distance ~10D
acceleration
sensors
air pressure and
temperature sens.
strain
gauges
op. data
tower
affected sectors
WT3
distance ~6D
foundation
n = 14 min-1
P = 413 kW
alpha = 5,5
Press Reset to
Reset
WT2
Figure 1: Measurement system installed at a
wind turbine
Additionally, operational data of the turbine is
measured simultaneously (electrical power, rotational speed, wind speed and azimuth position).
With this data, a reconstruction of the endured loads
before the beginning of the measurement campaign
is possible. As there are no permanent load records,
the turbine’s 10 min operational data is used for
correlation with the data of the measurement campaign.
The measurement is controlled by a robust mini PC
(operational from -20°C…+70°C, with no moving
parts), which can be remotely controlled by LAN or
UMTS. The data acquisition is performed by a special measurement LabVIEW-based software.
2.3 Reconstruction of endured loads
The goal of the measurement campaign is to determine the dynamic behaviour of the aged turbine,
and then to reconstruct the endured loads of its
WT1
distance ~8D
main wind direction SW
wind speed
in [m/s]
2005…2010
Figure 2: Sectoral wind speed distribution between 2005 and 2010 for the examined turbine
“WT2”, and wind sectors affected by the neighbouring turbines (note the location of WT1 in the
sector of main wind direction for WT2)
The data basis for the examined turbine is very
good, operational data is continuously available
from 2005 on with a 98% availability. The data include 10 min mean values, minima and maxima of
wind speed, rotor speed and electrical power, as
well as azimuth position and error status (times and
codes). The turbine is in good shape with regular
service inspections. However, significant differences
in the power and rotor speed curve are observed
compared to the neighbouring turbine of the nearly
similar type and age, figure 3, discussed later.
Paper for DEWEK 2015, Category: 9. Lifetime extension monitoring
power WT1
power WT2
rot. speed WT2
rot. speed [min-1]
el. power [kW]
rot. speed WT1
wind speed [m/s]
Figure 3: Electrical power and rotor speed versus mean wind speed, simultaneous 10 min operational
data of the examined wind turbine “WT2” and the almost identical turbine “WT1”
As research on the method is still in progress, the
measurement campaign is run for at least a year.
The data then allows comparing results for different
measurement periods and durations in order to
determine the reasonable minimum measurement
duration.
3.2 Alternative classing parameters
A basic problem for the classing of the operational
data is the lack of the turbulence intensity I or values that allow a direct calculation of I. The IEC
guideline [2] demands this value as a classing parameter, hence, a substitute turbulence intensity
value Isub is needed. In order to avoid additional
costs for simulated wind data for the site, it is investigated whether it is possible to derive Isub from the
turbine’s operational data at hand. Available operational data consists of mean, minimum and maximum values, measured on top of the nacelle, where
the rotor wake influences the measurement.
An approach investigated uses an established formula for the calculation of the extreme gust Ve50,
from mean wind speed Vm50 and turbulence intensity
I [1, 3, 4]:
Ve50 = Vm50 ∙ (1 + B ∙ I)
B is the gust factor, given as B = 2.8 in the literature
for various heights [4]. Rearranging this formula
while substituting Ve50 and Vm50 for the available
mean and max. values vmax,10 and vm,10 gives
Isub = (vmax,10 / vm,10 - 1) / B.
However, evaluations of various measurement data
have shown that the common value B = 2.8 can only
be reproduced when averaging wind speed bins
over longer periods. For 10 min time series, the
variance of vmax,10 is too large for a reasonable use
of a constant or site-calibrated value for B.
Although many combinations of minimum, maximum
and mean values from wind speed, power and rotor
speed have been investigated, no satisfying single
parameter has been identified yet.
3.3 Transferability to other turbines
During type certification [3, 5] it is assumed that the
prototype measurement results may be generalized
for the turbines of the same type. Hence, to reduce
measurement effort, it needs to be discussed
whether, it may be possible to use the load measurement data obtained at one aged turbine in order
to assess the endured loads and the residual service life of another turbine of the same turbine type
e.g. in the same wind farm.
At the site of the examined 600 kW turbine “WT2”,
the neighbouring “WT1”, figure 2, was erected at the
same time but has slightly different blades and in
addition other aerodynamic elements installed.
Furthermore, the number and frequency of the wind
sectors disturbed by fatigue-relevant wake effects
differs. Figure 4 shows that for WT1 the load spectra for undisturbed and disturbed wind sectors differ
significantly.
An analysis of the operational data of WT2 and WT1
() shows significant differences in the power and
rotor speed curves, figure 3. At the same wind
speed in partial load, WT1 has a higher power production and rotor speed than WT2. As the centrifugal mass forces and aerodynamic forces of the rotor
increase by the squared rotor speed, a corresponding increase in the general load levels seems possible for WT1. Furthermore, the frequency of time
periods where WT1 and WT 2 operate at their tower’s resonance frequency, which is within the operational range, will differ. However, it is also possible
that e.g. the wake-related increased loading on WT2
Paper for DEWEK 2015, Category: 9. Lifetime extension monitoring
600
disturbed sector
(06.11.-08.11.2014)
257° (south west)
vm = 5.4 m/s
Im = 11.7%
vibration range LC [mg]
500
400
300
200
undisturbed sector
(14.11.-16.11.2014)
152° (south east)
vm = 5.7 m/s
Im = 10.5%
100
0
1E+0
1E+1
1E+2
1E+3
1E+4
1E+5
load cycles N
Figure 4: Exemplary load spectra (lateral acceleration “LC”) for two day’s periods with approx. the same
mean wind speed vm, blue/rhombi: disturbed wind sector with wake effect and red/squares: undisturbed
wind sector where the loads are noticeably lower
led the OEM to a reduction of its power curve and
rotor speed curve, with the aim of reducing the load
level. Therefore, it seems that only further measurements at similar turbines will show whether or
not it is possible to use load spectra measured at
one turbine to assess the endured loads and remaining service life of another similar turbine. At
present, it seems to be necessary to perform individual load measurements.
4. Conclusion
With the presented project, a method for the realistic
assessment of the remaining service life of wind
turbine structures is developed and tested. The
approach can be seen as complementary to the
methods, proposed by the guidelines [1, 6]. Its main
innovation is the goal of a reliable, yet still economical assessment based on actual, individual load
measurement results. One challenge lies in the
correlation of the high resolution load measurement
data with the 10 min mean operational data which
do not contain directly the turbulence intensity,
which is the important second classing parameter.
Therefore, a reliable substitute value needs to be
derived. Further parameters influencing the classing
and thus the statistical distribution of the loads have
to be identified, with the goal to assure the acquisition of a representative amount of relevant data that
allows a realistic assessment of the remaining service life.
5. Acknowledgements
The doctoral stipend at TU Berlin is kindly granted
by the Reiner Lemoine Stiftung and supported by
BerlinWind GmbH. The measured wind turbine is
kindly provided by Terrawatt GmbH.
6. References
[1] Deutsches Institut für Bautechnik (DIBt): Richtlinie für Windenergieanlagen (Guideline for Wind
Turbines), 2012
[2] International Electrotechnical Commission: IEC
61400-13, Wind turbine generator systems –
Part 13: Measurement of mechanical loads,
2001
[3] Germanischer Lloyd (GL): Guideline for the
Certification of Wind turbines, 2010
[4] R. Gasch, J. Twele: Wind Power Plants, Springer, Second Edition, 2011
[5] International Electrotechnical Commission: IEC
61400-1, Wind turbines, Part1 1: Design requirements, 2005
[6] Germanischer Lloyd: Guideline for the Continued Operation of Wind Turbines, 2009